394 METALLURGY OF CAST IRON.
portion of the casting to be robbed of metal, which must be supplied, in order to prevent shrink-holes at all such points.
According to the view here presented, it will be also easy to understand that the resistance offered by the mould may often effect the expansion and shrinkage as well as the stibsequent contraction. Whether the power of expansion is as great as that of water in becoming frozen, is, as far as I know, undetermined. I do know that by casting between iron yokes or flask-ends, the longitudinal expansion of the bar may be prevented, as is seen in Test No. 9, Fig. 75, In such a case, of course, it is natural to suppose that the expansion must be in some other direction, and it may increase to a smaller degree the interior space necessary to be supplied with molten metal by feeding. The heat-conducting capacity of the mould, as determining the rate of solidification, may also effect the ap-„ parent result. Thus, a casting made in an "iron chill'' mould may show less shrinkage than if the same iron had been poured into a sand mould, because, in the latter case, the solidifying iron could have time and opportunity, by reason of the nature of the mould, to more expand it outward, thus increasing the interior space to be supplied with molten metal as already explained.
To return to the fact discovered by the writer, that hard grades of iron expand in solidifying more than soft grades, it may be said that this is contrary, not only to the general impressions, but also to the current explanation of the fact of expansion, which would ascribe it to the creation of graphitic carbon. If this were, the controlling cause, we should ex-r volume uf a easting, thereby causing the hottest or most fluidhe interior ofasonable to say, isU Citut